Pentagon labels AI firm Anthropic a provide chain danger ‘efficient instantly’ : NPR

[ad_1] <img src="https://npr.brightspotcdn.com/dims3/default/strip/false/crop/5000x3336+0+0/resize/1100/quality/50/format/jpeg/?url=http%3A%2F%2Fnpr-brightspot.s3.amazonaws.com%2Fd2%2Fb8%2Fdbc551884e5ab396b5f7ce397e94%2Fap26057633664886.jpg" data-template="https://npr.brightspotcdn.com/dims3/default/strip/false/crop/5000x3336+0+0/resize/{width}/quality/{quality}/format/{format}/?url=http%3A%2F%2Fnpr-brightspot.s3.amazonaws.com%2Fd2%2Fb8%2Fdbc551884e5ab396b5f7ce397e94%2Fap26057633664886.jpg" class="img" alt="Pages from the Anthropic website and the company's logo are displayed on a computer screen in New York on Thursday, Feb. 26, 2026." fetchpriority="high"/> Pages from the Anthropic web site and the corporate's brand are displayed on a pc display in New York on Thursday, Feb. 26, 2026. Patrick Sison/AP cover caption

toggle caption

Patrick Sison/AP The Trump administration is following by with its risk to designate synthetic intelligence firm Anthropic as a provide chain danger in an unprecedented transfer that might pressure different authorities contractors to cease utilizing the AI chatbot Claude.

The Pentagon stated in a press release Thursday that it has "formally knowledgeable Anthropic management the corporate and its merchandise are deemed a provide chain danger, efficient instantly."

The choice appeared to close down the chance for additional negotiation with Anthropic, practically per week after President Donald Trump and Protection Secretary Pete Hegseth accused the corporate of endangering nationwide safety.

Trump and Hegseth introduced a sequence of threatened punishments final Friday, on the eve of the Iran struggle, after Anthropic CEO Dario Amodei refused to again down over issues the corporate's merchandise could possibly be used for mass surveillance of Individuals or autonomous weapons.

Amodei stated in a press release Thursday that "we don't imagine this motion is legally sound, and we see no selection however to problem it in courtroom."

The Pentagon assertion stated, "this has been about one elementary precept: the navy having the ability to use know-how for all lawful functions. The navy won't permit a vendor to insert itself into the chain of command by proscribing the lawful use of a important functionality and put our warfighters in danger."

Amodei countered that the slim exceptions Anthropic sought to restrict surveillance and autonomous weapons "relate to high-level utilization areas, and never operational decision-making."

He stated there have been "productive conversations" with the Pentagon in latest days over whether or not it might maintain utilizing Claude or set up a "clean transition" if no settlement was reached. Trump gave the navy six months to part out Claude, which is already broadly embedded in navy and nationwide safety platforms. Amodei stated it is a precedence to verify warfighters will not be "disadvantaged of necessary instruments in the course of main fight operations."

Some navy contractors had been already reducing ties with Anthropic, a rising star within the tech business that sells Claude to a wide range of companies and authorities companies. Lockheed Martin stated it is going to "observe the President's and the Division of Battle's route" and look to different suppliers of enormous language fashions.

"We count on minimal impacts as Lockheed Martin isn't depending on any single LLM vendor for any portion of our work," the corporate stated.

How the Protection Division will interpret the scope of the danger designation is unclear. Amodei stated a notification Anthropic obtained from the Pentagon on Wednesday exhibits it solely applies to Claude's use by prospects as a "direct a part of" their navy contracts.

Microsoft stated its attorneys studied the rule and the corporate "can proceed to work with Anthropic on non-defense associated initiatives."

Pentagon attracts criticism for its choice The Pentagon's choice to use a rule designed to handle provide threats posed by international adversaries was met with broad criticism. Federal codes have outlined provide chain danger as a "danger that an adversary might sabotage, maliciously introduce undesirable operate, or in any other case subvert" a system with a purpose to disrupt, degrade or spy on it.

U.S. Sen. Kirsten Gillibrand, a New York Democrat and member of the Senate Armed Companies Committee and Senate Intelligence Committee, referred to as it "a harmful misuse of a software meant to handle adversary-controlled know-how."

"This reckless motion is shortsighted, self-destructive, and a present to our adversaries," she stated in a written assertion Thursday.

Neil Chilson, a Republican former chief technologist for the Federal Commerce Fee who now leads AI coverage on the Abundance Institute, stated the choice seems to be like "huge overreach that may damage each the U.S. AI sector and the navy's potential to accumulate one of the best know-how for the U.S. warfighter."

Earlier within the day, a bunch of former protection and nationwide safety officers despatched a letter to U.S. lawmakers expressing "severe concern" in regards to the designation.

"The usage of this authority in opposition to a home American firm is a profound departure from its meant goal and units a harmful precedent," stated the letter from former officers and coverage consultants, together with former CIA director Michael Hayden and retired Air Power, Military and Navy leaders.

They added that such a designation is supposed to "shield the USA from infiltration by international adversaries — from corporations beholden to Beijing or Moscow, not from American innovators working transparently underneath the rule of regulation. Making use of this software to penalize a U.S. agency for declining to take away safeguards in opposition to mass home surveillance and absolutely autonomous weapons is a class error with penalties that stretch far past this dispute."

Anthropic sees increase in shopper downloads Whereas dropping massive partnerships with protection contractors, Anthropic skilled a surge of shopper downloads over the previous week on account of individuals siding with its ethical stance. Greater than one million individuals signed up for Claude every day this week, the corporate stated, lifting it previous OpenAI's ChatGPT and Google's Gemini as the highest AI app in additional than 20 international locations in Apple's app retailer.

The dispute with the Pentagon has additionally additional deepened Anthropic's bitter rivalry with OpenAI that began when ex-OpenAI leaders, together with Amodei, began Anthropic in 2021.

Hours after the Pentagon punished Anthropic final Friday, OpenAI introduced a deal to successfully exchange Anthropic with ChatGPT in categorised navy environments.

OpenAI stated it sought comparable protections in opposition to home surveillance and absolutely autonomous weapons however later needed to amend its agreements, main CEO Sam Altman to say he should not have rushed a deal that "appeared opportunistic and sloppy."

Amodei additionally expressed remorse about his personal half in that "tough day for the corporate," saying Thursday he wished to "straight apologize" for an inner notice he despatched to Anthropic employees that attacked OpenAI's conduct and urged Anthropic was being punished for not giving "dictator-like reward" to Trump.